Академический Документы
Профессиональный Документы
Культура Документы
Optimization in Design
Progress Report
t
c
sweep angle
Motivation
Typical Engineer’s Method
• Establish requirements.
• Try to optimize by
changing one (or
several) design variable
at a time.
F2
Not perfect curve.
• Objective functions have many local minima
(artifact of numerical procedures).
• The graph implies that we need more work on
optimization.
Using other optimization codes
• To investigate alternative formulations, we need to use
tools that are external to FLOPS.
• NPSOL (Stanford Software, Gill et al.) is a set of Fortran
subroutines for minimizing a smooth function subject to
bounds on variables, linear constraints and smooth
nonlinear constraints.
• It uses a sequential quadratic programming (SQP)
algorithm.
• Call previous Matlab codes to adjust input variables,
perform analysis and read output results.
• Use NPSOL to minimize the result (weighted objective
function)
• Used “out of the box”, NPSOL did not
provide better results than FLOPS itself
– Price of running FLOPS is quite high
– May not be efficient enough in handling this
special problem
– May need fine tuning
• A bootstrapping strategy of the two
codes can do quite well
Minimization Obtained
Minimum of Gross Weight
FLOPS 213554
NPSOL 221495
FLOPS + NPSOL 211920
2 * (FLOPS + 210046
NPSOL)
Progress
• Unconstrained optimal
design of subsonic aircraft.
– Done using Mathematica’s
FindMinimum command and
FLOPS
– “Optimal” designs are often
unrealistic (because of the
problem formulation).
“Optimal”
• Understand relationships
between aircraft design and 1
overpressure signal.
0
-1 -0.5 0 0.5 1